Raid Ledger

Warcraft Logs stores a combatant-info snapshot at the start of every boss pull: each player's equipped items with enchant, temporary enchant and gem ids, and the auras on them at that moment. Raid Ledger resolves those ids against an offline item, enchant and gem database built from the wowsims project's tooltip dumps, plus an extra.json file for Anniversary-only items the dump predates. An item id the database does not know is never silently dropped: the shadow resistance audit lists it in Flags as "Unknown items: [ids]", and the enchants and gems audit treats an unknown off-hand item as one that needs an enchant, although it cannot know the item's socket count and does not mark the row. The audits therefore describe the state at each pull, not between pulls, and a player who died before a snapshot's pull can show fewer buffs than they actually used.

Two things are not in the log at all: race and talent ranks. Draenei and Undead get +10 shadow resistance, and a Mage's Magic Absorption talent gives +2 resistance to everything per point (up to 5). Today these come from two seed files carried over from the Python tool; the race file is keyed by character name, and a Name-Server key also works. Alliance Shamans are inferred as Draenei and Horde Paladins as Blood Elf from class and faction, where the faction comes from the report's guild (or from a character probe stored with the snapshot), so a report with no guild attached gets no inference. Everyone else without an entry shows ? and a flag. A per-guild settings page is the planned replacement.

For the selected fight, each raider gets a Gear SR (items, assumed green suffixes, enchants, armor kits, glyphs and resist-all gems) and a Buffed SR (gear plus the strongest protection buff, flask, Broiled Bloodfin, Gift of Arthas, Mage or Demon Armor, racial and talent). The targets are buffed at or above 244 and gear at or above 160; green means at target, red means below, and tanks are never colored. Each row also shows Race / Class and Role. The table is sorted by gear SR so the weakest links are at the bottom.

Buffs in one group do not stack, so only the strongest counts: Prayer of Shadow Protection at +70 beats Gift of the Wild at +25 in the protection group, and the flask, food, Gift of Arthas and armor groups each count once. The buff columns are headed Shadow Protection, Flask, Broiled Bloodfin, Gift of Arthas and Mage (the armor column, which shows n/a for any class other than Mage or Warlock). Hover a buff cell to see which buff was actually found. Random-suffix greens (the "of Shadow Protection" pieces most people wear for this fight) are a special case because the log does not store the suffix. Every random-suffix item is counted as its shadow-resistance roll: known items use a Wowhead-verified table, unlisted ones are interpolated from same-slot greens at that item level and listed in Flags as "Estimated green suffix value", and items with no SR suffix at all count 0. The engine notes every assumed green as "ASSUMED SR suffix — verify in game", but the site does not show per-item notes, so a green with a known suffix value produces no visible flag; check the Greens column instead.

The enchants and gems audit answers one question per slot: should it carry an enchant, and does it. Head, shoulder, back, chest, wrist, hands, legs, feet and main hand always need one. Off hand needs one only if it is a weapon or shield: a one-hand weapon in the off hand is required, a Held In Off-hand item is n/a, and an item the offline database does not know is treated as required. Ranged needs one only if it is a bow, gun or crossbow, and neck, waist, trinkets, shirt and tabard never do. Socket counts come from the offline database, so an empty socket is caught even though the log only records filled ones. An enchant id the database does not know is counted as present, and a log that shows more gems than the item has sockets is noted in the engine; neither note is shown on the site.

Each row shows Enchants done/needed, Gems filled/sockets, a Status of OK or MISSING, and a Missing column with the exact problem ("Legs: no enchant", "Head: 2 empty sockets"). Rings follow the auto policy: ring enchants exist only for Enchanters, so both rings are required only for a player who already has one ring enchanted. A separate Warnings column lists things that do not change Status: a poor, common or uncommon quality gem, a meta socket filled with a non-meta gem, and no temporary weapon enchant (oil, sharpening stone, poison or shaman imbue) on the main hand at the pull.

The consumables audit covers every boss encounter in the report rather than one pull. For each player it shows Fights (how many of the covered fights they attended), Flask / elixirs as the average uptime of whichever was up (a flask, a battle elixir or a guardian elixir; under the "any" policy a flask or a single elixir is a complete setup), Food as the average Well Fed uptime shown as a percentage (the engine resolves which food was eaten, but the site does not display the name), and Potions as used / possible. The budget is one potion per two minutes of fight plus the opener, because every TBC potion shares a two-minute cooldown. A Potion breakdown lists what was used (for example 4× Haste Potion, 2× Super Mana Potion). Rows are grouped Tanks, Healers, DPS. A player who appears only on trash and attended no boss fight is left out.

Uptime cells are green at 90 percent or better and red under 50 percent. Potion cells are green at 50 percent of budget or better and red under 25 percent. Flags spell it out: "No flask or elixir in n/N fights" with the boss names, which fires for any fight where the best flask or elixir uptime was under 50 percent, "No food in n/N fights", "Food dropped below 90% in n fights (died / ate late)", "0 potions in n/N boss fights", "Potion usage x/y possible (z%)" when under half the budget, and "Unverified classification" when a flask, elixir or food buff was recognized only by its name rather than a known spell id (potions, healthstones, drums and runes matched by name are counted without that flag). The engine also has an "Only one elixir, no flask ... counts as 50% coverage" flag, but it only fires under the both policy, which the site does not run. Healthstones, drums and runes are tracked in the engine but not yet shown as columns.

Trash pulls have no aura snapshot, so they are off by default. Include trash carries each player's flask, elixir and food state forward from the last boss pull or event, capped at the buff's duration (flask two hours, elixirs one hour, food thirty minutes), until the next boss snapshot resets it. Potions used on trash never count against the budget; they appear as "(+n on trash)" in the breakdown. Treat trash numbers as approximate.

Click any column header to sort, click again to flip. Sorting is value-aware: "17/21" sorts by the ratio and "72%" by the number. The first click on a number column sorts descending, which puts "?" and blanks at the bottom; the second click flips to ascending and they move to the top. Text columns such as Player and Flags start ascending. The Flags, Missing, Warnings and Potion breakdown columns wrap as text so a long list stays readable. Every audit lives at a plain URL, so sharing a result is pasting a link into Discord; whoever opens it sees the same cached tables without signing in, unless the report is private.

Every fetch from Warcraft Logs is stored once in the RaidKit database and shared by all viewers. A report that ended less than 24 hours ago is treated as live: opening it re-checks the fight list against the API, a fight whose end time moved is refetched, and the consumables lines are rebuilt when new fights appear. Older reports are served from cache until someone asks for a refetch. The Refresh button on the report page adds ?refresh=1 to the report URL, which refetches only the report header (title, fight list, visibility); the audit links it renders carry no refresh flag, so a cached audit is still served unless its fight's end time or the set of fights changed. To refetch an audit itself, add ?refresh=1 to that audit's own URL. Signed in, /ledger lists the reports you have opened (most recently opened first, up to 12) and the homepage the last 4; the list is yours alone — no one else's reports appear in it and yours appear in no one else's — and a Forget button drops a report from it. Signed out, nothing is listed.

Public and unlisted reports are read with RaidKit's own Warcraft Logs API client, exactly as Warcraft Logs would serve them to anyone with the link, and the site finds the report across the www, classic and fresh partitions on its own. A private report is different: it is read only with the token of a signed-in person who connected a Warcraft Logs account that can see it, and cached data for a private report is served only after Warcraft Logs confirms on that request that the viewer's token still opens it. The cache never makes a private report more public than it is on Warcraft Logs.

When a fetch fails, the notice says the next step for you specifically: Sign in, Connect Warcraft Logs, or Reconnect Warcraft Logs when a link has expired. Long-lived tokens are refreshed automatically near expiry; a refused refresh asks you to reconnect. What RaidKit keeps for the connection is your Warcraft Logs username and the tokens it issued. Disconnecting deletes them. A RaidKit account itself is created with Google, GitHub or an email address and password (with email verification, forgot and reset password, and an optional authenticator app as a second factor); Warcraft Logs shares no email address, so Continue with Warcraft Logs only signs in an account that already connected it from Settings.

The audit engine is a line-for-line TypeScript port of the author's wow-audit Python tool, and parity tests hold both implementations to the same numbers on the same fixtures (for example one character at 174 gear and 244 buffed shadow resistance, matched against the in-game character sheet). The Python tool remains the offline reference and still has a few things the web version does not: an editable Magic Absorption rank picker and per-audit Excel workbooks. The web table's Talent cell already carries the rank-picker data (player, talent, rank, maximum), but the site does not render a picker yet.